Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

REVIEW

Form follows function: the genomic organization of cellular differentiation

Steven T. Kosak1 and Mark Groudine1,2,3

1Division of Basic Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109, USA; 2Department of Radiation Oncology, University of Washington School of Medicine, Seattle, Washington 98195, USA

The extent to which the nucleus is functionally orga- regarding its function. The current paradigm of gene nized has broad biological implications. Evidence sup- regulation includes the binding of site-specific transcrip- ports the idea that basic nuclear functions, such as tran- tion factors, the recruitment of cofactors and general scription, are structurally integrated within the nucleus. transcription factors, and the incorporation of multiple Moreover, recent studies indicate that the linear arrange- modifications to both the DNA and the histones that ment of genes within eukaryotic genomes is nonrandom. organize it (Felsenfeld and Groudine 2003). This descrip- We suggest that determining the relationship between tion of transcription belies its enormous complexity, fu- nuclear organization and the linear arrangement of genes eled by an ever-increasing catalog of proteins dedicated will lead to a greater understanding of how transcrip- in one way or another to its regulation. Additionally, tomes, dedicated to a particular cellular function or fate, evidence supporting the role of nuclear localization in are coordinately regulated. Current network theories transcriptional regulation indicates that it is insufficient may provide a useful framework for modeling the inherent to know the components of transcription (Francastel et complexity the functional organization of the nucleus. al. 2000). Rather, a thorough understanding of the pro- cess requires knowing its functional organization within the nucleus. In this sense, transcription should not be Louis Sullivan, whose early efforts helped pioneer the viewed simply as a process that turns on a specific gene, development of the skyscraper, is considered one of the but as a process that governs within the genome an en- most important architects of the last century. However, tire network of genes (a transcriptome) that gives rise to it is his dictum—“form ever follows function”—for a particular cellular function or fate (such as cell divi- which he is perhaps best known. Just as this imperative sion, differentiation, or apoptosis). Therefore, the chal- has influenced generations of architects, the idea that lenge is to uncover the nuclear organization of gene ac- structure reflects function provides a useful perspective tivity and to determine whether genomes are specifically for a biologist’s view of the cell. In many ways, the struc- structured. ture and function of cellular and subcellular organelles The form DNA takes in the nucleus is a result of at are inseparable; that is, disruptions in organelle function least three prevailing components, its organization into can lead to perturbations in its structure. Upon inhibi- chromatin, the linear order of genes and repetitive ele- tion of rRNA transcription, for example, the nucleolus ments along their respective chromosomes, and the spa- becomes disordered and ultimately disappears (Leung tial localization of genes and repeats within the nucleus. and Lamond 2003). This integration of structure and cel- Current efforts with molecular, cell biological, and ge- lular function allows for conservation of resources and nomic approaches are attempting to elucidate the role facilitates regulation at multiple levels. each of these components of DNA plays in regulating Although a completely sequenced genome may repre- nuclear processes. Clearly, the forms of DNA permissive sent a genetic blueprint, molecular biologists currently for gene transcription and gene silencing are of particular lack a key with which to fully grasp how this sequence is importance. This review will survey what is currently related to the development and subsequent maintenance known about the localization of genes spatially within of a given organism. Following Sullivan’s example, a the nucleus and linearly in the genome, focusing on how comprehensive understanding of genomic sequence may these organizational states may help facilitate the or- require considering its arrangement in the nucleus; the chestrated gene expression that results in cellular differ- form DNA takes in the nucleus reveals not only its entiation. Finally, the review will explore how this co- higher-order structure, but it may impart information ordinated expression may be modeled by current net- work theories. [Keywords: Nuclear structure; genomic organization; transcription; expression neighborhood; genetic networks] 3Corresponding author. Spatial organization of gene activity within the nucleus E-MAIL [email protected]; FAX (206) 667-5894. Article and publication are at http://www.genesdev.org/cgi/doi/10.1101/ Cellular differentiation is generally accompanied by co- gad.1209304. ordinated changes in gene expression and alterations in

GENES & DEVELOPMENT 18:1371–1384 © 2004 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/04; www.genesdev.org 1371 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine nuclear structure. For example, as genes are silenced, the of snRNAs by base pairing and aligning the modifying extent of chromatin condensation often increases, and enzymes (Darzacq et al. 2002). Importantly, fibrillarin is extended regions of DNA are packaged into heterochro- structurally similar to methyltransferases, and mutation matin (John 1988). The amount and distribution of con- of its yeast ortholog results in unmethylated pre-rRNA densed chromatin is similar in differentiated cells of the and a loss of ribosomes in the cytoplasm (Tollervey et al. same lineage, but the pattern varies in the nuclei of dif- 1991; Wang et al. 2000). Given these characteristics, the ferent cell types (Leitch 2000). These types of observa- Cajal body is likely a site of snoRNP and snRNP pro- tions have led to the idea that nuclear organization may cessing. Therefore, the IGC and the Cajal body reveal be cell-type-specific; the topological organization of the that the machinery involved in the processing of tran- interphase nucleus may reflect the differentiated state of scripts is organized within the nucleus. Although these the cell and may be involved in the establishment and nuclear structures do not actually participate in the propagation of tissue-specific patterns of gene expression mechanism of transcription, they provide evidence for (Francastel et al. 2000). Although we are just beginning the emerging idea that transcription, splicing, and fur- to understand the link between the form and function of ther transcript modification are integral (Bentley 2002) chromatin in the interphase nucleus, technological ad- and may therefore be spatially organized within the vances, such as fluorescence in situ hybridization (FISH) nucleus (Fig. 1). and multidimensional (3-D and 4-D) fluorescence mi- PML nuclear bodies (PML-NBs) were first observed as croscopy, have already led to a greater understanding of a result of the t(15;17) translocation detected in nearly all the functional organization of the nucleus. cases of acute promyelocytic leukemia (APL). This trans- location involves the fusion of two genes, PML and the retinoic acid receptor ␣ (RAR␣), and leads to the disrup- The form of transcriptional activity tion of PML-NBs (Ruggero et al. 2000). Although the ac- tual mechanistic function of the PML-NB has remained Nuclear bodies elusive, it has been implicated in many fundamental cel- The nucleolus, whose function and components are lular processes, including transcription (Zhong et al. known in detail, is the most well-characterized nuclear 2000). In support of a role in transcription, regulatory body and is exclusively reviewed elsewhere (e.g., Leung factors such as cyclic AMP response-binding protein and Lamond 2003). The idea that other nuclear processes (CBP) and retinoblastoma protein (pRB) have been shown may occur in discrete sites within the nucleus owes an to interact directly with PML (Ruggero et al. 2000). A initial example to Hewson Swift (see Acknowledg- recent localization analysis of gene-dense domains in the ments). Using electron microscopy, Swift described the human genome revealed that transcriptionally robust nonhomogenous fine structure of the nucleus, identify- loci are associated with PML-NBs (Wang et al. 2004). ing areas of low-electron density into which fibrils ex- Although RNA FISH and RNA interference (RNAi) ex- tend and associate with small, dense structures, now periments did not reveal a strict correlation of transcrip- known as interchromatin granule clusters (IGCs; Swift tional activity and localization to PML-NBs, the major- 1959; Lamond and Spector 2003). Evidence that IGCs ity of RNA transcript foci analyzed were positioned at contain RNA and colocalize with transcribed genes led the bodies. Furthermore, treatment of an APL cell line to the idea that they may represent sites of active tran- with all-trans retinoic acid resulted in the generation of scription (Lamond and Spector 2003). However, immu- PML-NBs and the concomitant localization of an active nofluorescence microscopy with antibodies against com- locus to the emergent structures (Wang et al. 2004). It is ponents of the splicing apparatus (such as snRNPs and therefore plausible that PML-NBs play a role in tran- SR proteins, pre-messenger RNA splicing factors with scription, possibly as a reservoir of concentrated tran- characteristic arginine–serine repeats) has revealed that scriptional regulators (Fig. 1). Given that they also con- IGCs most likely serve as a reservoir of proteins involved tain proteasomes, PML-NBs may play a further regula- in mRNA processing (Huang and Spector 1996). Local- tory role by degrading transcription factors in response to ization of active genes near IGCs may therefore facilitate external stimuli. transcription by providing concentrations of splicing components. For example, the association of SR proteins The interchromatin compartment with IGCs is modulated by phosphorylation, and over- expression of an SR kinase results in dissipation of IGCs The nuclear bodies described above indicate an inherent and a concomitant reduction in pre-mRNA splicing tendency toward organization of the transcription and (Sacco-Bubulya and Spector 2002). transcript-processing machinery. Moreover, the per- Similarly, the Cajal body, whose protein components turbation of function associated with each body results include coilin and fibrillarin, colocalizes with a subset of in the loss of its nuclear form. The existence of such active genes and is disrupted by perturbation of RNA structures raises the question of where they (and the polymerase II (Pol II) transcription or protein translation processes they support) are positioned in the nucleus (Ogg and Lamond 2002). However, immunofluorescence relative to their substrate, chromatin. Euchromatin de- analyses have revealed that the Cajal body contains fines the gene-rich, transcriptionally active (or potenti- snRNPs and snoRNPs (Gall 2000). Furthermore, Cajal bod- ated) portion of the genome. Futhermore, euchromatin ies contain guide RNAs that facilitate the modification has a characteristic histone modification pattern and is

1372 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Genomic organization of cellular differentiation

Figure 1. The nuclear organization of gene activity and gene silencing. This schematic representation indicates what is currently known of how active and po- tentiated genes are positioned within the nucleus and how silenced genes are com- partmentalized in repressive subcompart- ments. See text for details. further defined by nucleosomes with specific histone CTs more often than gene-poor domains (Mahy et al. variants, such as H3.3 (Vermaak et al. 2003). FISH analy- 2002). These results suggest that CTs, as determined by sis with combined locus-specific probes and whole-chro- whole-chromosome paints, may represent the relatively mosome paints has permitted the visualization of genes more condensed domains of a chromosome. A gene or relative to their chromosome territories (CTs), the dis- multigene locus in a state of open chromatin modifica- crete structures that individual chromosomes form in tion and structure may therefore be excluded from the the interphase nucleus (Fig. 1). Initial studies revealed CT when visualized by FISH. Regardless, the looping of that genes are preferentially positioned at territory sur- a gene array from its CT may increase its association faces, whereas intergenic DNA is found within the CT with the nuclear bodies that facilitate transcription and (Zirbel et al. 1993; Kurz et al. 1996). These observations transcript processing. led to the idea that an intervening compartment runs Analysis of the endogenous wild-type and derivative throughout the nucleus in the space between the dis- mutant ␤-globin gene loci has helped to clarify the sig- crete CTs, creating an interchromosome domain en- nificance of looping of a locus from its CT (Ragoczy et al. riched for the nuclear bodies involved in transcription 2003). In erythroid cells, the ␤-globin locus is looped and splicing (Cremer et al. 1993). More recent studies away from its CT at a high frequency prior to transcrip- have confirmed that transcription does occur at the tional induction. Thus, looping is not a consequence of surface of CTs, but that this surface runs throughout the transcription per se, but may also represent a poised state invaginated contours of a territory, creating an inter- prior to activation. However, in the absence of the locus chromatin compartment (Verschure et al. 1999; Visser et control region (LCR), which is required for the high-level al. 2000). Therefore, the interchromatin compartment globin gene transcription induced upon terminal differ- (IC) model predicts that active genes are organized at entiation, the locus is positioned at the CT surface. the continuous surfaces of CTs to facilitate their regula- Furthermore, if the ␤-globin LCR is replaced by se- tion by bringing them into proximity with the nuclear quences from the B-cell-specific immunoglobulin heavy bodies positioned in the IC (Cremer and Cremer 2001; chain (IgH)3ЈC␣ LCR, an element that represses tran- Fig. 1). scription of reporters in non-B cells (Madisen and Grou- In addition to being at the surface or the interior of a dine 1994), looping is partially restored, but is now CT, a third type of territorial position has recently correlated with localization of the looped locus to peri- emerged, cell- and activity-dependent organization of centromeric heterochromatin (PCH) in another chromo- multigene loci in a large loop (several megabase pairs) some territory. Interestingly, the IgH locus is looped emanating from the CT. For example, it has been re- from its CT specifically in pro-B cells (where it is tran- ported that active loci consisting of coordinately regu- scriptionally active), but is not positioned near hetero- lated genes, such as the major histocompatibility and chromatin. These results argue against a simple correla- epidermal differentiation complexes, are looped away tion of elevated transcriptional activity and looping from the central body of the CT during robust transcrip- away from CTs; rather, extrusion from the CT may play tion (Volpi et al. 2000; Williams et al. 2002). In addition, a significant role in cell-type-specific transcriptional ac- gene-rich domains, with generally ubiquitous expression tivation or repression of a locus by localizing it to a par- patterns, have a propensity to be looped away from their ticular position within the nucleoplasm.

GENES & DEVELOPMENT 1373 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine

The form of transcriptional repression terns of particular cytokines expressed during Th1 ver- sus Th2 differentiation can be explained by association Heterochromatin of the inactivated genes with PCH (Grogan et al. 2001). Silenced genes possess a distinct chromatin configura- These studies suggest that organization of genes into tion and are specifically compartmentalized into tran- heterochromatin can lead to transcriptional repression; scriptionally repressive nuclear subcomartments however, they do not demonstrate the means by which (Felsenfeld and Groudine 2003). Staining of interphase this silencing is achieved (Fig. 1). nuclei with DNA-intercalating dyes reveals that the Analyses of the native ␤-globin locus and derivative nucleus is organized into regions of weak and intense transgenes in erythroid cells have provided a direct link labeling, which correlate with euchromatic (active) and between PCH association and gene activity. For ex- heterochromatic (inactive) chromatin domains, respec- ample, when linked to a reporter gene integrated at PCH, tively. Heterochromatin is further classified as either an erythroid-specific enhancer (5ЈHS2) derived from the constitutive heterochromatin (CH) or facultative hetero- ␤-globin LCR confers localization of the transgene away chromatin (FH). Although the exact structure of CH has from PCH and stable reporter expression (Francastel et yet to be characterized, it demonstrates a regular nucleo- al. 1999). Analysis of wild-type and mutant human ␤-glo- somal spacing (as opposed to euchromatin and FH) and is bin loci have also shed light on the role of PCH associa- refractory to DNase I and endonuclease enzymes, indi- tion and gene activity. In erythroid cells, the wild-type cating a highly organized and condensed structure (Dil- locus is located away from PCH, and displays an active lon and Festenstein 2002). Furthermore, CH is highly chromatin structure as assayed by nuclease sensitivity methylated, gene poor, late replicating, and transcrip- and histone H3 and H4 hyperacetylation (Schübeler et al. tionally repressive (Wallrath 1998; Bridger and Bickmore 2000). In contrast, a ␤-globin locus carrying a large natu- 1998). CH is comprised of arrays of tandem repeats (or rally occurring deletion encompassing the LCR and 35 satellites), whereas FH is a consequence of euchromatin kb upstream (␥␦␤° thalassemia) colocalizes with PCH being packaged into a condensed, transcriptionally re- and adopts an inactive chromatin structure as revealed pressive structure during cellular development (Dillon by nuclease insensitivity and histone H3 and H4 hypo- and Festenstein 2002). Therefore, CH represents a cell- acetylation, a state resembling the inactive wild-type type-independent organization of chromatin, whereas ␤-globin locus in lymphocytes (Brown et al. 2001). FH is lineage dependent and actively formed. Underscor- The studies discussed above have revealed a correla- ing these elemental differences, a recent study has dem- tion between the activity of a gene and its proximity to onstrated that FH in erythrocytes can form in the ab- CH. In all of the cases, the genes that are ultimately sence of heterochromatin protein 1 (HP1), which is a sequestered at heterochromatin are significant in the dif- requisite component of CH (Gilbert et al. 2003). ferentiation of the involved cell type. That is, the genes Although much remains to be determined about the localized to PCH are oftentimes the genes whose sup- structure and function of CH, its role in gene regulation pression is necessary in that cell type, or in that stage of has been well documented. Pericentromeric heterochro- cell development. It is likely, then, that PCH-association matin (PCH) describes the less-homogenous regions of facilitates the formation of a repressive chromatin struc- satellite DNA adjacent to true centromeres, which may ture and that the examples described above are, in effect, localize to the periphery of CH clusters (Lundgren et al. facultative heterochromatin. The active recruitment to 2000). In Drosophila, the bwD allele exerts its domi- PCH may therefore be reserved for those genes that must nance over the wild-type locus by forcing its association be silenced for differentiation to occur, or those whose with PCH (Csink and Henikoff 1996). The large insertion regulation must be modulated to ensure the precise de- of satellite DNA in the bwD allele causes it to organize velopmental progression. with similar repeats found in heterochromatin, and so- An analysis of transgenes comprised of copies of ␭-5 (a matic pairing serves to recruit the wild-type allele to this gene involved in B-cell development) integrated into repressive domain. In murine-developing B cells, the PCH, underscores the significance of CH association in lymphocyte-restricted transcriptional regulator Ikaros cell development (Lundgren et al. 2000). Despite integra- colocalizes with PCH through direct DNA binding tion into PCH and localization to the outside of CH clus- (Brown et al. 1997; Cobb et al. 2000). T-cell-specific ters, the transgenes demonstrate position effect variega- genes and developmentally regulated B-cell genes asso- tion (PEV) in pre-B cells, indicating that proximity to CH ciate with heterochromatic Ikaros clusters specifically does not preclude activity. However, loss of a potent HS when they are inactive (Brown et al. 1997). Ikaros ap- site results in the internalization of the transgene into pears to associate with binding sites in a gene’s regula- CH in fibroblasts (cells in which the gene is inactive) and tory element, and then recruits the gene to PCH through in the reduction of expression in pre-B cells, although Ikaros-binding sites found in CH (Trinh et al. 2001). Fur- remaining at the surface of CH. The position of the ⌬HS thermore, a study of the immunoglobulin gene (Ig) loci transgene in pre-B cells may reflect the availability of in immature B cells has demonstrated a nonrandom as- regulatory proteins that directly impact its activity, as a sociation with PCH, which may have implications in the genetic background heterozygous for EBF (a gene re- allelic exclusion occurring at these loci (Skok et al. quired for early B-cell development) results in the inter- 2001). In developing T cells, which derive from a com- nalization of the ⌬HS transgene into CH and to a signifi- mon progenitor as B cells, the specific expression pat- cant reduction in its activity. Therefore, the formation of

1374 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Genomic organization of cellular differentiation facultative heterochromatin at PCH may lead to the pro- periphery in the regulation of these intricately regulated gressive silencing of genes that are obligately repressed gene arrays (Kosak et al. 2002). In lymphoid progenitors for cellular differentiation, reflecting the need to localize (as well as embryonic stem cells), the inactive IgH and silenced genes in a particular nuclear subcompartment Ig␬ loci are sequestered at the nuclear periphery. During to preclude their response to the fluctuating concentra- early B-cell development, both loci are relocalized to the tions of regulatory proteins. Interestingly, the FH that nuclear center, which may represent a transcriptionally forms during terminal erythroid differentiation coin- permissive nuclear environment. Localization to the cides with a large-scale relocation of proteins associated nuclear center is not necessarily a function of transcrip- with gene repression (e.g., MeCP2, HDAC1, and MafK) tion, as the centrally positioned Ig␬ loci are not active. from CH to other nuclear subcompartments, reflecting Interestingly, the IgH locus undergoes compaction the large-scale nuclear condensation that occurs at this (wherein distal ends of the 3-Mbp array colocalize, im- developmental stage (Francastel et al. 2001). plying a looped structure) when it is centrally located in Modifications to the N termini of histones can regu- the nucleus and is poised to undergo long-range V(D)J late the binding of proteins involved in chromatin orga- recombination. A null mutation of the interleukin-7 re- nization and gene regulation (Felsenfeld and Groudine ceptor ␣, which results in a block early in B-cell devel- 2003). HP1 is perhaps the best-understood protein in- opment, abrogates relocalization of the loci from the pe- volved in the transcriptional repression by heterochro- riphery and prevents the compaction of the IgH locus. matin, having been shown to localize to CH clusters and A recent study has further delineated the steps involved to mediate gene silencing (Eissenberg and Elgin 2000). in the compaction of the locus, indicating that the B- Studies of HP1 have shed light on a potential mechanism cell regulatory protein Pax-5 in conjunction with an for the maintenance and spreading of repressive hetero- unknown B-cell-specific factor may induce the close chromatin. Importantly, histone H3 methylated at Lys 9 juxtaposition of the ends of the IgH array (Fuxa et al. specifically binds to HP1 (Bannister et al. 2001; Lachner 2004). et al. 2001). Furthermore, this association is dependent FISH analysis of whole chromosomes in human nuclei on the activity of histone methyltransferases (HMTs) has revealed that a gene-poor chromosome (18) is prefer- that specifically modify histone H3 on Lys 9 (Rea et al. entially localized to the nuclear periphery, whereas a 2000). Because HP1 and the HMTs are colocalized in gene-rich chromosome (19) is more centrally disposed in heterochromatin domains, these results suggest a means the nucleus (Croft et al. 1999). This preferential associa- for transcriptionally repressive chromatin structures to tion is maintained, even in the context of a balanced be maintained as well as spread to adjacent cis se- translocation between the two chromosomes, with the quences. In addition, it is also possible that this mecha- translocated portions of 18 and 19 residing peripherally nism may function in trans, silencing genes brought to and centrally, respectively. Further analysis of gene- heterochromatin domains by heterochromatin-associat- dense and gene-poor chromosomes has confirmed the ing proteins, like Ikaros (Fig. 1). tendency for gene-poor chromosomes to be positioned at the nuclear periphery (Boyle et al. 2001). Cross-species analysis has revealed that this behavior is not restricted Nuclear periphery to the human nucleus (Tanabe et al. 2002). As described Despite early indications that transcription may be lo- below, gene-rich chromosomal domains are the most calized to the nuclear periphery (Hutchison and Wein- highly expressed regions of the human genome; there- traub 1985) and a recent demonstration that boundary fore, the demonstrations of gene-rich chromosomes or- activities (BAs, which protect the expression status of ganized into the nuclear center may simply be a reflec- active domains) involve the tethering of active chroma- tion of their overall level of activity. tin to the nuclear pore complex (NPC; Ishii et al. 2002), The studies described above strongly suggest that the the nuclear periphery has primarily been demonstrated nuclear periphery may represent a transcriptionally re- to represent a transcriptionally repressive nuclear com- pressive nuclear compartment distinct from CH, which partment. The nuclear periphery’s role in repression has often resides in perinuclear clusters. For example, the been well established in budding yeast. A number of peripheral localization of silent Ig loci does not involve studies have collectively demonstrated that yeast telo- association with PCH (Kosak et al. 2002). In fact, this meres form clusters at the nuclear periphery, which study indicates that the nuclear lamina itself may play a leads to an enrichment of the Sir proteins known to be role in the sequestration and inactivity of perinuclear involved in gene silencing (Cockell and Gasser 1999). loci. The major components of the nuclear lamina are The ability of this peripheral compartment to repress the lamins, type-V intermediate filament proteins that transcription was tested in a study in which a reporter polymerize to form the lamin network that is juxtaposed gene was tethered to the nuclear envelope. Making use of to the inner nuclear membrane of the nuclear envelope. a Gal4–DNA-binding domain/integral membrane pro- There are two classes of lamins, A type and B type. Ex- tein fusion, a reporter flanked by Gal4-binding sites was pression of the A-type lamins is developmentally regu- inducibly repressed in a Sir-dependent manner (Andrulis lated, whereas the B-type lamins are ubiquitously ex- et al. 1998). pressed (Mounkes et al. 2003). The nuclear lamina, An analysis of the murine Ig loci during lymphocyte through lamin B, interacts directly with DNA and chro- development has shown the involvement of the nuclear matin, as well as indirectly through lamin-binding pro-

GENES & DEVELOPMENT 1375 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine teins (Gotzmann and Foisner 1999). In addition, proteins Chromosome organization demonstrated to be involved in gene silencing have also Beyond the movement of individual genetic loci, there is been shown to associate with the lamina, including HP-1 evidence that chromosomes may themselves be mobile. (Kourmouli et al. 2000). Therefore, although much Two recent studies utilized an H2B–GFP fusion protein works need to be done before a causative effect in gene and photobleaching to analyze the overall order of chro- silencing can be attributed to localization at the nuclear mosomes through the cell cycle. In an analysis of HeLa periphery, growing evidence supports the idea that it rep- cells, chromosomes were shown to maintain their local- resents a transcriptionally repressive nuclear subcom- ization in daughter cells in approximately half the nuclei partment (Fig. 1). studied (Walter et al. 2003). Furthermore, chromosomes The relationship between structure and function in were shown to be mobile during the early Gap 1 cell the nucleus is clearly evidenced by mutations in the A- cycle (G1). A similar analysis that modeled a random and type lamin gene (LMNA, with major splicing variants A nonrandom organization of chromosomes to be expected and C), resulting in several human diseases, collectively from the photobleaching experiment showed that the or- termed “laminopathies.” These diseases include muscu- ganization of chromosomes is significantly nonrandom, lar dystrophy, cardiomyopathy, partial lipodystrophy, or maintained, during mitosis (Gerlich et al. 2003). De- and progeria syndromes (Genschel and Schmidt 2000; spite the discrepancies between these results, they both Mounkes et al. 2003). Of particular interest is how mu- argue that an inherent chromosome organization may tation of a single gene that is broadly expressed in differ- exist that is remembered upon cell division. In support of entiated tissues could result in several tissue-specific the suggestion of a defined chromosomal organization in disease phenotypes. One possible explanation is that la- the nucleus, studies of the chromosomes and gene loci min A/C, localized at the nuclear periphery as well as involved in translocations that lead to leukemia have internal, perinucleolar foci, establishes a structure in dif- revealed a propensity for translocation partners to be spa- ferentiated cells on which transcriptional regulators and tially proximal (Parada et al. 2002; Roix et al. 2003). their respective target genes are organized. In support of These results argue for a functional organization of the this idea, lamin A/C has been found to interact with genome at the level of the chromosome. The exact na- transcription factors, such as pRb and SREBP1, impor- ture of this organization, and whether the organization tant in the differentiation of mesenchymal tissues, particular to a given cell type is altered as the cell re- which are most affected by mutations in LMNA (Man- sponds to external stimuli or, in fact, differentiates, has cini et al. 1994; Lloyd et al. 2002). yet to be determined.

Linear arrangement of gene activity within the genome Chromatin mobility If the nuclear localization of a gene is involved in its The organization of the transcriptional machinery and regulation and the development of any cell type requires the compartmentalization of silenced genes suggest that the regulation of hundreds and perhaps thousands of genes must be mobile within the nucleus to be appropri- genes (its transcriptome), then cellular differentiation ately positioned. Currently, the study of chromatin mo- may be accompanied by large-scale nuclear reorganiza- bility has yielded conflicting results in the comparison of tion. Regardless of whether this reorganization is due to human and yeast nuclei. In humans, small movements a direct or indirect mechanism, the individual relocal- of 0.5 µM have been demonstrated, which allows a gene ization of hundreds of spatially distant genes would be to sample a very small fraction of the total nuclear vol- inefficient. Therefore, it is possible that an underlying ume (Chubb et al. 2002). However, in light of the IC linear order of genes along chromosomes exists to facili- model, these small movements may be sufficient to lo- tate the coordinated regulation of the transcriptome. The calize a gene to a relevant nuclear body or repressive idea that coregulated genes may share linear positions compartment. Also, use of the lac operator-repressor sys- within a genome has two obvious precedents; namely, tem (which allows visualization of chromatin through the operons of prokaryotic genomes (as well as that of arrays of lac-binding sites) revealed that a late-replicat- Caenorhabditis elegans) and the gene arrays of eukary- ing, heterochromatic domain undergoes large-scale otes, both of which have been instrumental in the un- movement from the nuclear periphery to the interior derstanding of transcriptional regulation. For example, prior to replication (Li et al. 1998). In yeast (and Dro- the lineage-restricted IgH and ␤-globin loci are examples sophila), 0.5 µM movements have also been detected, of gene arrays that share a common genomic position but given the significant difference in nuclear size, these and are intricately regulated in specific cell types. Fur- movements permit a gene to travel upward of half the thermore, as discussed above, it has also been shown nuclear diameter (Gasser 2002). Furthermore, the move- that both of these loci have nuclear localization patterns ment of loci in yeast has been shown to be energy de- that parallel their state of activity. Although these two pendent, unlike the small-scale movements in humans. gene arrays (and some others that have been character- Despite the differences between human and yeast, the ized) are the result of duplication events, it is neverthe- evidence for short, diffusional movements are compat- less likely that coregulated genes unrelated in sequence ible with the role of nuclear localization in gene regula- homology may be organized in linear clusters through- tion. out the genome.

1376 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Genomic organization of cellular differentiation

The advent of multiple genome-wide analytical tech- plotted along the linear gene order of the chromosomes niques has provided the means to explore genomes for (Cohen et al. 2000). An analysis of these maps with ex- networks of unique genes that are clustered within the pression data obtained from cell cycle phases, sporula- genome and involved in a common cellular function or tion, and the pheromone response, revealed an inherent in the differentiation of a particular lineage. If such lin- organization of the yeast genome; a highly significant ear gene clusters exist, they would support a model in percentage of nonduplicated, coexpressed genes are adja- which coregulated genes exhibit physical proximity cent (and to a lesser degree form triplets) along the chro- along their chromosomes to facilitate their regulation. mosomes. Furthermore, these adjacent genes also tend to Evidence from all species so far examined has revealed be functionally related. To ascertain the nature of the that genomes are nonrandomly organized (Fig. 2). coregulation of adjacent genes, their regulatory se- quences were examined. Although adjacent genes do not necessarily have similar UASs, there are several ex- Yeast amples in which one of the adjacent genes lacks a UAS (Cohen et al. 2000), indicating that adjacency may allow As discussed above, budding yeast has provided an ex- neighboring genes to share a single regulatory element. cellent model for the study of nuclear localization affect- ing gene activity, specifically in the repressive nature of peripheral localization. Telomere position effect (TPE), Worm which is caused by the tethering of telomeres at the pe- riphery amid the localized enrichment of repressive SIR Unlike other eukaryotes, which have not been shown to proteins, provides an important example of how a gene’s contain operons, as much as 25% of C. elgans’ coding linear position within the genome can affect its regula- sequence may be organized into polycistronic operons of tion (Hediger and Gasser 2002). An analysis of chromo- two to eight genes (Blumenthal 1998). Clearly, operons some correlation maps of Saccharomyces cerevisiae has exemplify coregulation through proximal positioning revealed that beyond TPE, there is an underlying order to within the genome. The multiple genes of an operon the yeast genome. Correlation maps allow the expres- share a common regulatory domain, thereby ensuring sion patterns from various conditions or cell stages to be the coexpression of genes typically involved in a com-

Figure 2. Eukaryotic genomes are nonrandomly organized. Evidence for a shared linear organization of similarly expressed genes has been uncovered in each of the species analyzed. Although the nature of the organization is not necessarily the same among the different species, the fact that all of the genomes do possess localized activity argues for its significance in gene regulation. See text for details on each organism.

GENES & DEVELOPMENT 1377 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine mon function. Recently, however, an organization of in- bined protein in the case of the Ig loci. These arrays are dividual, monocistronic genes has been uncovered found both for tissue-specific genes, as well as ubiqui- within the worm genome (Roy et al. 2002). mRNA tag- tously expressed genes. Therefore, there is a precedent ging [which makes use of an epitope-tagged poly(A)-bind- for the proximal positioning of genes that share a com- ing protein] was developed for the isolation of tissue- mon function, even though the example of arrays prima- specific transcripts from whole larvae. After excluding rily indicate duplication events. Evidence also indicates genes within an operon and tandem duplications, an that there may be a broader organization to tissue-re- analysis of muscle-specific genes revealed that they are stricted transcriptomes in the mouse. For example, an clustered together in groups of two to five throughout examination of ESTs from extraembryonic tissues from the genome (Roy et al. 2002). Unlike yeast, however, post-implantation mouse embryos (d7.5) revealed an or- these clustered genes do not necessarily share a common ganization of 155 cDNA clones localized in clusters on cellular function. Morever, analysis of microarray data subregions of chromosomes 2, 7, 9, and 17 (Ko et al. from germ-line cells showed that sperm, oocytes, and the 1998). Although the potential clustering of these genes germ line itself demonstrate an organization of tissue- was not tested, these data indicate that there is a non- specific expression similar to that of muscle (Roy et al. random distribution of coregulated genes at the level of 2002). These data and those from yeast suggest that tran- the chromosome. The t-complex itself, located on chro- scriptomes, dedicated to a cell state or to a particular cell mosome 17, represented 6.5% of all cDNAs. Similarly, type, exist in organized centers or neighborhoods, and an analysis of expression profiles from embryonic, neu- that changes in expression patterns correspond with a ronal, and hematopoietic stem cells revealed the t-com- shift in the genomic organization of the transcriptome. plex to be enriched for shared stem-cell genes (Ramalho- Santos et al. 2002). Examination of the differentiation of a hematopoietic progenitor into erythroid and neutrophil Fly cell types indicates an organization of transcriptomes into adjacent, coregulated genes that changes upon dif- Drosophila is widely known for its polytene chromo- ferentiation (S.T. Kosak, D. Scalzo, F. Li, S. Hall, T. En- somes, an arrangement of the genome in salivary glands ver, and M. Groudine, in prep.). in which numerous rounds of replication without mito- sis result in enormous polyploid chromosomes. Polytene chromosomes have been shown to exhibit puffs in re- Human gions of high levels of transcription, encompassing do- An integration of the human genomic sequence with mains of presumably coregulated genes (Thummel SAGE (serial analysis of gene expression) data for ge- 2002). Polytene puffs may therefore represent a physical nome-wide mRNA expression patterns from 12 tissue manifestation of clustered genes with similar expression types, has provided a Human Trasnscriptome Map patterns. A microarray analysis of expression from Dro- (Caron et al. 2001) that reveals the human genome is sophila determined under 80 different experimental con- nonrandomly organized into regions of high and low lev- ditions, has revealed an organization of nonhomologous, els of gene activity. The highly active regions, RIDGEs coexpressed genes in groups of 10–30, covering between (regions of increased gene expression), are separated by 20 and 200 kbp (Spellman and Rubin 2002). Although large regions of low activity (antiridges, not unlike val- these genes demonstrate coregulation, shared functions leys). Importantly, RIDGES and valleys coincide with for genes in each group was not established. Importantly, gene-dense and gene-poor chromosomal domains, re- the grouped genes show highly correlated levels of ex- spectively. Therefore, gene activity is inherently com- pression, suggesting that the domain organization is a partmentalized along the chromosome, which is analo- reflection of an active chromatin structure that stretches gous to the further subdivision of lineage-restricted or through the region. An analysis of expressed sequence coregulated genes being clustered in the genomes of the tags (ESTs) databases has also demonstrated the cluster- model organisms described above. RIDGES also demon- ing of genes within the Drosophila genome; however, strate a high GC content, SINE density, and a low-intron this examination focused on the tissue-specific expres- length (Versteeg et al. 2003), implying a higher-order or- sion profiles from the testis, head-region, and embryo ganization of the genome that may be a reflection of (Boutanaev et al. 2002). In each cell type, the coregulated chromosomal structure and/or a strategy for gene regu- genes were found to be significantly organized into clus- lation. Analysis of the linear organization of the mouse ters of three or more genes, with a trend toward large genome has reveled a nonhomogenous distribution simi- groupings. Therefore, the clustering of coregulated, lin- lar to that of human, indicating that this type of genomic eage-restricted genes indicates a functional organization pattern has been conserved (Mural et al. 2002; S.T. Ko- of transcriptomes that define a given cell type. sak, D. Scalzo, F. Li, S. Hall, T. Enver, and M. Groudine, in prep.). A recent analysis of SAGE data indicates that RIDGEs may be a consequence of the population of these Mouse regions by the ubiquitously and highly expressed house- Vertebrate genomes have many well-characterized loci keeping genes and that tissue-restricted transcriptomes that encompass gene arrays that demonstrate coregula- have a tendency to be clustered (Lercher et al. 2002). tion and a shared function, or even an ultimately recom- Also, paralleling the evidence from the worm, analysis of

1378 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Genomic organization of cellular differentiation a genomic transcript map of human skeletal muscle et al. 2002). An obvious possibility is that proximal genes genes revealed that genes expressed in this lineage are share enhancer elements. In the case of yeast and C. el- concentrated on three chromosomes (17, 19, and X) in egans, the sharing of a common regulatory element may in five chromosomal regions (Bortoluzzi et al. 1998). There- fact occur at adjacent genes (Cohen et al. 2000; Lercher et fore, in addition to the overall genomic organization of al. 2002). The evidence from other eukaryotes, however, RIDGEs, there is a further level of organization of lin- suggests that a more general effect on transcriptional eage-specific genes. regulation may be at work. One possible explanation of There have been many indications, such as operons this effect is that an increased local concentration of and position effect, that genomes are not homogenously regulatory sequences, which are identical or involved in organized. Now, from the genomic approaches described the regulation of related genes, create a hub of the pro- above, it appears that there is an elemental nonrandom teins that, in turn, bind these sequences. Specifically, the organization of eukaryotic genomes. Coexpressed genes grouping of genes may decrease the effective off-rate of demonstrate a propensity to be adjacent or grouped along regulatory proteins through the localization of binding the genome. These gene clusters can be functionally re- sites. This is an attractive possibility, given that numer- lated or involved in the transcriptome of a specific cell ous FRAP studies have indicated a high-diffusion con- type. These latter features offer evidence that there may stant for both regulatory and structural nuclear proteins be evolutionary constraints upon the genomic organiza- (Phair and Misteli 2000; Cheutin et al. 2003). The high tion of coexpressed genes. The profound effect on cellu- mobility of regulatory proteins (such as transcription fac- lar differentiation of a single translocation giving rise to tors) is particularly significant, as a given binding site is leukemia offers evidence of the importance of the regu- found in many locations throughout the genome that are latory consequence of a gene’s chromosomal context not germane to gene regulation (Bulyk 2003; Fig. 3A). (Rowley 1998). Therefore, the link between expression The study of lymphopoiesis has provided several im- and position strongly indicate a role in gene regulation portant examples in support of the functional relevance (although other processes, including splicing and repli- of regulatory protein concentrations. Study of the Pax-5 cation, may also be related to the genome’s overall orga- transcription factor, which can act as a transcriptional nization). Further analysis will be necessary to deter- repressor and activator, revealed that at lower nuclear mine whether an inherent organization of the nucleus concentrations, the protein maintains its positive regu- exists that reflects the nonrandom linear arrangement of latory role due to a 20-fold higher affinity of activator genes, and whether this nuclear organization is altered elements (Wallin et al. 1998). Additionally, as discussed during differentiation. above, localized concentrations of the EBF transcription The actual role clustering plays in gene regulation re- factor can override the suppressive effects of PCH mains to be established. Nevertheless, the available data (Lundgren et al. 2000). Finally, the overall level of the suggest several potential mechanisms for how expres- transcription factor PU.1 in progenitor populations can sion neighborhoods may influence gene regulation (Oliver affect the hematopoietic developmental pathway, with

Figure 3. Potential roles for genetic clus- tering in gene regulation. (A) The clustering of genes may facilitate gene regulation by the formation of expression neighborhoods (or hubs) through a feedback of increased binding sites and a subsequent concentra- tion of the regulatory proteins that bind them. Such a model would require the spa- tial association of these chromosome re- gions within the nucleus. The chromosome regions may include the alleles from ho- molgous chromosomes (as depicted), or other regions that have clusters of similarly regulated genes. (B) As opposed to the ac- tive mechanism in A, the clustering of gene expression may be the result of a domain architecture of chromatin. In this model, a potent regulatory element or the additive spreading of a chromatin configuration may lead to the coexpression of adjacent genes. It is likely that boundary elements would be involved to prevent the spreading of this effect, as well as to prevent the encroachment of silencing mechanisms. The degree of influence identified for these domains is similar to the extent of chromosome territory looping, and it is interesting to consider whether the domain architecture may reflect the spatial positioning of blocks of chromatin into a transcriptionally permissive nuclear subcompartment (see The Interchromatin Compartment).

GENES & DEVELOPMENT 1379 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine high levels resulting in myeloid cells and low levels in B tion describes a structure that forms from molecular in- cells (DeKoter and Singh 2000). Therefore, creating an teractions in a steady state. The idea that self-organiza- expression neighborhood through the localization of tion may describe the mechanism that forms the nuclear regulatory elements would be an efficient means to po- bodies and subcompartments (e.g., heterochromatin) is tentiate regional gene activity (Fig. 3A). based largely on FRAP analysis (Misteli 2001b). These Another possibility for the role of linear arrangement studies demonstrate the rapid diffusion of both regula- in gene regulation, which is not mutually exclusive with tory and structural proteins in the nucleus. Because they protein concentrations, is that a potent regulatory ele- are in constant flux, random interactions of proteins are ment (or elements) may influence the expression status thought to seed the formation of transient structures; in within a chromosomal region. This enhancer may di- other words, a stable structure is achieved by the dy- rectly activate the individual promoters of the adjacent namic, continuous exchange of its components. Self-or- genes, or it may simply lead to the spreading of histone ganization may therefore explain how the functional in- modifications that, in turn, would affect the transcrip- teraction of ribosomal proteins, rRNA transcription fac- tional status of surrounding genes (Fig. 3B). Evidence tors, and the rDNA template lead to the genesis of the from the fly does not support a spreading effect, however, nucleolus (Misteli 2001a). Interestingly, the introduction as there does not appear to be a gradual decline of influ- of rDNA into ectopic sites within the genome leads to ence the further from the center of the expression neigh- the formation of micronucleloi around the integrated borhood (surrounding genes are instead either on or off; genes. As this example illustrates, the concept of self- Spellman and Rubin 2002). In fact, the Drosophila data organization holds promise in facilitating our under- suggests the formation of a static domain, perhaps standing of the structural organization of cellular func- through the use of insulators, that delimits the local ef- tion. Self-organization, however, does not address the fect on expression. A recent genomic analysis of HP1 and nonrandom nature of the components of a functional Su(var)3–9 binding supports a domain architecture of structure. Specifically, the underlying order of eukary- gene expression, as developmentally regulated genes dis- otic genomes argues against a random association of a play uniform patterns of association with one or both of gene and its requisite regulatory machinery. Form may the proteins (Greil et al. 2003). Either by a spreading of indeed follow function, but the form of a self-organized modifications or the establishment of a domain, it is structure may be predisposed by a nonrandom organiza- interesting that looping from CTs (as described above) tion of its parts. appears to correlate with the range of effect seen at the In recent years, graph (or network) theory, the math- genomic level (Ragoczy et al. 2003). It is possible that CT ematical field that explores how networks form, has looping may be a physical manifestation of the potenti- made considerable progress in analyzing the nature of ated or activated state of an expression neighborhood. real-world networks (Barabási and Oltvai 2004). Net- Therefore, it will be very interesting to determine work theory has, until recently, been dominated by the whether looped domains colocalize to repressive sub- idea that networks are inherently random in their for- compartments (like PCH) or to regions permissive for mation and consequent organization. The interconnec- transcription (like the nuclear center; Fig. 3B). tions of a random network, defined by nodes (the enti- ties) and links (the connections), follow a Poisson distri- bution, with the vast majority of nodes having a Cellular differentiation as a genetic network common, relatively small number of links and rare out- The evidence reviewed above indicates that the nuclear lier nodes with many more or fewer links. The study of localization of a particular gene can significantly affect real-world networks, however, has revolutionized the its activity and that a shared characteristic of eukaryotic field, revealing that random networks do not predomi- genomes is the nonrandom organization of genes related nate in the natural world. Initial analysis of the World by shared expression patterns and possibly function. Wide Web and the Internet determined that these net- This spatial and linear organization of genes suggests a works do not follow a Poisson distribution, but rather, highly structured nucleus, in which form and function they best fit a power-law degree distribution (P(k) ∼ k−␥). are intricately linked. However, as discussed above, an The diminishing tail of the power-law curve gave these inherent organization of chromosomes for a particular networks their name, scale-free, which refers to the lack progenitor cell type, which changes upon cellular differ- of a prevalent linkage number. Importantly, all scale-free entiation, has yet to be demonstrated. Furthermore, a networks so far examined demonstrate a ␥ (degree expo- spatial analysis of all the active genes within a nucleus nent) between 2 and 3, which is influenced by the very has not been performed. Therefore, if we are to approach few nodes that have a tremendous number of links. the concept of a nonrandom, functionally coordinated Scale-free networks have two primary rules for their for- genome, it is necessary to model the functional structure mation and maintenance, a scale-free network expands of the nucleus. continuously with the addition of new nodes, and these The concept of self-organization has been put forth to nodes are added preferentially to sites that are already explain the behavior of the nucleus (in addition to other well connected (Barabási and Albert 1999). These prin- organelles; Misteli 2001a). As opposed to a self-assembly ciples explain the hubs (the highly connected nodes) seen mechanism, in which constituent proteins form a static in real-world networks (Fig. 4A). In essence then, scale- nuclear structure in a state of equilibrium, self-organiza- free networks describe a kind of self-organization

1380 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Genomic organization of cellular differentiation

tein network has uncovered that those proteins with the greatest number of links are those most likely to be le- thal (Jeong et al. 2001), an organization of genes within the genome and in the nucleus may indicate a mecha- nism by which proximity facilitates regulation. The proximal positioning of coregulated genes, described above, provide just the nodes that would permit model- ing the genetic organization of cellular differentiation as a scale-free network. Although not elaborated in the de- scription of expression neighborhoods, the majority of coregulated genes are not necessarily grouped within the genome. As demonstrated in the Drosophila genome, however, there are chromosome domains with a prepon- derance of coexpressed genes. These regions, in a scale- free model, would represent the hubs of genetic activity, as so many genes are contained in a confined genomic region (e.g., Featherstone and Broadie 2002). If the model is allowed to include the association of allelic regions of homologous chromosomes, then these hubs are even more pronounced in their local gene activity. Further- more, if other highly expressed regions may also com- municate in the coordinate activity of these domains, a significant regional neighborhood of expression would result (Fig. 4B). We propose that there may be evolutionary constraints upon the disruption of localized gene expression, whether its origin is due to duplication events or other- wise, which have ensured that coregulated genes in- Figure 4. Network theory and genomic organization. (A) The volved in a common function maintain a shared linear recent description of numerous real-world networks has trans- position within the genome. By spatially restricting the formed network theory. “Scale-free” describes a special aspect position of genes, their regulation can be coordinated of real-world networks, the organized centers, or hubs, of highly linked nodes. This behavior has been shown in metabolic as through a concentration of regulatory proteins or by the well as proteomic pathways, and is depicted here playing a role spreading of chromatin modifications and activation me- in genetic interactions. It is based upon the constant growth of diated by enhancers. Further analysis will be required to real-world networks, and the preferential attachment of those verify that gene clustering truly facilitates coordinate added nodes. (B) The nonrandom organization of the linear and gene regulation and determine whether this linear orga- spatial genome suggests features of scale-free networks. The nization is reflected in the spatial organization of coregu- localized position of multiple coregulated genes would facilitate lated in the nucleus. In addition, whether the nonran- their intercommunication by helping to form and then utilize dom order of genes on chromosomes necessitates a par- localized concentration of regulatory proteins. An expression ticular nuclear organization of chromosomes remains to hub, or neighborhood, of this kind may include any two alleles be established. (chromosomes of the same color represent homologs), with their mirrored linear gene order, in addition to regions of other chromosomes with clustered genes undergoing similar regula- Acknowledgments tion. We acknowledge the many contributions of the late Hewson Swift to cell biology and to our understanding of the nucleus. In no small way, the substance of this review is indebted to him. that is instructed by its inherent substructure. These We apologize to those researchers whose work we were not able qualities make network theory an engaging model with to include due to size constraints. We also thank our colleagues which to approach the genomic organization of the tran- in the Groudine lab for useful discussions, and Mike Bulger, scriptional regulation of differentiation. Scale-free mod- Steve Henikoff, Dave Scalzo, Bill Schubach, Erica Smith, and els have, in fact, already been used to describe metabolic Jon Soderholm for critical reading of this manuscript. S.K. is a and proteomic networks in biological systems (Jeong et fellow of the Jane Coffin Childs Memorial Fund for Medical al. 2000, 2001; Giot et al. 2003). Research. This work has been aided by a grant from The Jane In essence, a network is an exchange of information Coffin Childs Memorial Fund for Medical Research and was that obligates either a physical interaction or the sharing supported by NIH grants (DK44746 and HL57620) to M.G. of regulatory information. The coordinate regulation of hundreds of genes therefore describes a network that References guarantees the intercommunication of the integral, yet Andrulis, E.D., Neiman, A.M., Zappulla, D.C., and Sternglanz, individually regulated genes that affect a cell response or R. 1998. Perinuclear localization of chromatin facilitates cellular differentiation. Just as analysis of the yeast pro- transcriptional silencing. Nature 394: 592–595.

GENES & DEVELOPMENT 1381 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine

Bannister, A.J., Zegerman, P., Partridge, J.F., Miska, E.A., Tho- compartmentalization of the cell nucleus. Cold Spring Harb. mas, J.O., Allshire, R.C., and Kouzarides, T. 2001. Selective Symp. Quant. Biol. 58: 777–792. recognition of methylated lysine 9 on histone H3 by the HP1 Croft, J.A., Bridger, J.M., Boyle, S., Perry, P., Teague, P., and chromo domain. Nature 410: 120–124. Bickmore, W.A. 1999. Differences in the localization and Baraba´si, A.L. and Albert, R. 1999. Emergence of scaling in ran- morphology of chromosomes in the human nucleus. J. Cell. dom networks. Science 286: 509–512. Biol. 145: 1119–1131. Baraba´si, A.L. and Oltvai, Z.N. 2004. Network biology: Under- Csink, A.K. and Henikoff, S. 1996. Genetic modification of het- standing the cell’s functional organization. Nat. Rev. Genet. erochromatic association and nuclear organization in Dro- 5: 101–113. sophila. Nature 381: 529–531. Bentley, D. 2002. The mRNA assembly line: Transcription and Darzacq, X., Jady, B.E., Verheggen, C., Kiss, A.M., Bertrand, E., processing machines in the same factory. Curr. Opin. Cell. and Kiss, T. 2002. Cajal body-specific small nuclear RNAs: A Biol. 14: 336–342. novel class of 2Ј-O-methylation and pseudouridylation guide Blumenthal, T. 1998. Gene clusters and polycistronic transcrip- RNAs. EMBO J. 21: 2746–2756. tion in eukaryotes. Bioessays 20: 480–487. DeKoter, R.P. and Singh, H. 2000. Regulation of B lymphocyte Bortoluzzi, S., Rampoldi, L., Simionati, B., Zimbello, R., Barbon, and macrophage development by graded expression of PU.1. A., d’Alessi, F., Tiso, N., Pallavicini, A., Toppo, S., Cannata, Science 288: 1439–1441. N., et al. 1998. A comprehensive, high-resolution genomic Dillon, N. and Festenstein, R. 2002. Unravelling heterochroma- transcript map of human skeletal muscle. Genome Res. tin: Competition between positive and negative factors regu- 8: 817–825. lates accessibility. Trends Genet. 18: 252–258. Boutanaev, A.M., Kalmykova, A.I., Shevelyov, Y.Y., and Nur- Eissenberg, J.C. and Elgin, S.C. 2000. The HP1 protein family: minsky, D.I. 2002. Large clusters of co-expressed genes in Getting a grip on chromatin. Curr. Opin. Genet. Dev. the Drosophila genome. Nature 420: 666–669. 10: 204–210. Boyle, S., Gilchrist, S., Bridger, J.M., Mahy, N.L., Ellis, J.A., and Featherstone, D.E. and Broadie, K. 2002. Wrestling with pleiot- Bickmore, W.A. 2001. The spatial organization of human ropy: Genomic and topological analysis of the yeast gene chromosomes within the nuclei of normal and emerin-mu- expression network. Bioessays 24: 267–274. tant cells. Hum. Mol. Genet. 10: 211–219. Felsenfeld, G. and Groudine, M. 2003. Controlling the double Bridger, J.M. and Bickmore, W.A. 1998. Putting the genome on helix. Nature 421: 448–453. the map. Trends Genet. 14: 403–409. Francastel, C., Walters, M.C., Groudine, M., and Martin, D.I. Brown, K.E., Guest, S.S., Smale, S.T., Hahm, K., Merken- 1999. A functional enhancer suppresses silencing of a trans- schlager, M., and Fisher, A.G. 1997. Association of transcrip- gene and prevents its localization close to centrometric het- tionally silent genes with Ikaros complexes at centromeric erochromatin. Cell 99: 259–269. heterochromatin. Cell 91: 845–854. Francastel, C., Schubeler, D., Martin, D.I., and Groudine, M. Brown, K.E., Amoils, S., Horn, J.M., Buckle, V.J., Higgs, D.R., 2000. Nuclear compartmentalization and gene activity. Nat. Merkenschlager, M., and Fisher, A.G. 2001. Expression of ␣- Rev. Mol. Cell Biol. 1: 137–143. and ␤-globin genes occurs within different nuclear domains Francastel, C., Magis, W., and Groudine, M. 2001. Nuclear re- in haemopoietic cells. Nat. Cell. Biol. 3: 602–606. location of a transactivator subunit precedes target gene ac- Bulyk, M.L. 2003. Computational prediction of transcription- tivation. Proc. Natl. Acad. Sci. 98: 12120–12125. factor binding site locations. Genome Biol. 5: 201. Fuxa, M., Skok, J., Souabni, A., Salvagiotto, G., Roldan, E., and Caron, H., van Schaik, B., van der Mee, M., Baas, F., Riggins, G., Busslinger, M. 2004. Pax5 induces V-to-DJ rearrangements van Sluis, P., Hermus, M.C., van Asperen, R., Boon, K., and locus contraction of the immunoglobulin heavy-chain Voute, P.A., et al. 2001. The human transcriptome map: gene. Genes & Dev. 18: 411–422. Clustering of highly expressed genes in chromosomal do- Gall, J.G. 2000. Cajal bodies: The first 100 years. Annu. Rev. mains. Science 291: 1289–1292. Cell Dev. Biol. 16: 273–300. Cheutin, T., McNairn, A.J., Jenuwein, T., Gilbert, D.M., Singh, Gasser, S.M. 2002. Visualizing chromatin dynamics in inter- P.B., and Misteli, T. 2003. Maintenance of stable heterochro- phase nuclei. Science 296: 1412–1416. matin domains by dynamic HP1 binding. Science 299: 721– Genschel, J. and Schmidt, H.H. 2000. Mutations in the LMNA 725. gene encoding lamin A/C. Hum. Mutat. 16: 451–459. Chubb, J.R., Boyle, S., Perry, P., and Bickmore, W.A. 2002. Chro- Gerlich, D., Beaudouin, J., Kalbfuss, B., Daigle, N., Eils, R., and matin motion is constrained by association with nuclear Ellenberg, J. 2003. Global chromosome positions are trans- compartments in human cells. Curr. Biol. 12: 439–445. mitted through mitosis in mammalian cells. Cell 112: 751– Cobb, B.S., Morales-Alcelay, S., Kleiger, G., Brown, K.E., Fisher, 764. A.G., and Smale, S.T. 2000. Targeting of Ikaros to pericen- Gilbert, N., Boyle, S., Sutherland, H., de Las Heras, J., Allan, J., tromeric heterochromatin by direct DNA binding. Genes & Jenuwein, T., and Bickmore, W.A. 2003. Formation of facul- Dev. 14: 2146–2160. tative heterochromatin in the absence of HP1. EMBO J. Cockell, M. and Gasser, S.M. 1999. Nuclear compartments and 22: 5540–5550. gene regulation. Curr. Opin. Genet. Dev. 9: 199–205. Giot, L., Bader, J.S., Brouwer, C., Chaudhuri, A., Kuang, B., Li, Cohen, B.A., Mitra, R.D., Hughes, J.D., and Church, G.M. 2000. Y., Hao, Y.L., Ooi, C.E., Godwin, B., Vitols, E., et al. 2003. A A computational analysis of whole-genome expression data protein interaction map of Drosophila melanogaster. Sci- reveals chromosomal domains of gene expression. Nat. ence 302: 1727–1736. Genet. 26: 183–186. Gotzmann, J. and Foisner, R. 1999. Lamins and lamin-binding Cremer, T. and Cremer, C. 2001. Chromosome territories, proteins in functional chromatin organization. Crit. Rev. Eu- nuclear architecture and gene regulation in mammalian karyot. Gene Expr. 9: 257–265. cells. Nat. Rev. Genet. 2: 292–301. Greil, F., van der Kraan, I., Delrow, J., Smothers, J.F., de Wit, E., Cremer, T., Kurz, A., Zirbel, R., Dietzel, S., Rinke, B., Schrock, Bussemaker, H.J., van Driel, R., Henikoff, S., and van E., Speicher, M.R., Mathieu, U., Jauch, A., Emmerich, P., et Steensel, B. 2003. Distinct HP1 and Su(var)3–9 complexes al. 1993. Role of chromosome territories in the functional bind to sets of developmentally coexpressed genes depending

1382 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Genomic organization of cellular differentiation

on chromosomal location. Genes & Dev. 17: 2825–2838. affects changes in higher order chromatin structure associ- Grogan, J.L., Mohrs, M., Harmon, B., Lacy, D.A., Sedat, J.W., ated with activation of a heterochromatic gene. Cell and Locksley, R.M. 2001. Early transcription and silencing of 103: 733–743. cytokine genes underlie polarization of T helper cell subsets. Madisen, L. and Groudine, M. 1994. Identification of a locus Immunity 14: 205–215. control region in the immunoglobulin heavy-chain locus Hediger, F. and Gasser, S.M. 2002. Nuclear organization and that deregulates c-myc expression in plasmacytoma and Bur- silencing: Putting things in their place. Nat. Cell. Biol. kitt’s lymphoma cells. Genes & Dev. 8: 2212–2226. 4: E53–E55. Mahy, N.L., Perry, P.E., and Bickmore, W.A. 2002. Gene density Huang, S. and Spector, D.L. 1996. Dynamic organization of pre- and transcription influence the localization of chromatin mRNA splicing factors. J. Cell Biochem. 62: 191–197. outside of chromosome territories detectable by FISH. J. Hutchison, N. and Weintraub, H. 1985. Localization of DNAase Cell. Biol. 159: 753–763. I-sensitive sequences to specific regions of interphase nuclei. Mancini, M.A., Shan, B., Nickerson, J.A., Penman, S., and Lee, Cell 43: 471–482. W.H. 1994. The retinoblastoma gene product is a cell cycle- Ishii, K., Arib, G., Lin, C., Van Houwe, G., and Laemmli, U.K. dependent, nuclear matrix-associated protein. Proc. Natl. 2002. Chromatin boundaries in budding yeast: The nuclear Acad. Sci. 91: 418–422. pore connection. Cell 109: 551–562. Misteli, T. 2001a. The concept of self-organization in cellular Jeong, H., Tombor, B., Albert, R., Oltvai, Z.N., and Barabasi, architecture. J. Cell. Biol. 155: 181–185. A.L. 2000. The large-scale organization of metabolic net- ———. 2001b. Protein dynamics: Implications for nuclear ar- works. Nature 407: 651–654. chitecture and gene expression. Science 291: 843–847. Jeong, H., Mason, S.P., Barabasi, A.L., and Oltvai, Z.N. 2001. Mounkes, L., Kozlov, S., Burke, B., and Stewart, C.L. 2003. The Lethality and centrality in protein networks. Nature laminopathies: Nuclear structure meets disease. Curr. Opin. 411: 41–42. Genet. Dev. 13: 223–230. John, B. 1988. The biology of heterochromatin. Cambridge Uni- Mural, R.J., Adams, M.D., Myers, E.W., Smith, H.O., Miklos, versity Press, Cambridge. G.L., Wides, R., Halpern, A., Li, P.W., Sutton, G.G., Nadeau, Ko, M.S., Threat, T.A., Wang, X., Horton, J.H., Cui, Y., Pryor, E., J., et al. 2002. A comparison of whole-genome shotgun-de- Paris, J., Wells-Smith, J., Kitchen, J.R., Rowe, L.B., et al. rived mouse chromosome 16 and the human genome. Sci- 1998. Genome-wide mapping of unselected transcripts from ence 296: 1661–1671. extraembryonic tissue of 7.5-day mouse embryos reveals en- Ogg, S.C. and Lamond, A.I. 2002. Cajal bodies and coilin—Mov- richment in the t-complex and under-representation on the ing towards function. J. Cell. Biol. 159: 17–21. X chromosome. Hum. Mol. Genet. 7: 1967–1978. Oliver, B., Parisi, M., and Clark, D. 2002. Gene expression Kosak, S.T., Skok, J.A., Medina, K.L., Riblet, R., Le Beau, M.M., neighborhoods. J. Biol. 1: 4. Fisher, A.G., and Singh, H. 2002. Subnuclear compartmen- Parada, L.A., McQueen, P.G., Munson, P.J., and Misteli, T. talization of immunoglobulin loci during lymphocyte devel- 2002. Conservation of relative chromosome positioning in opment. Science 296: 158–162. normal and cancer cells. Curr. Biol. 12: 1692–1697. Kourmouli, N., Theodoropoulos, P.A., Dialynas, G., Bakou, A., Phair, R.D. and Misteli, T. 2000. High mobility of proteins in Politou, A.S., Cowell, I.G., Singh, P.B., and Georgatos, S.D. the mammalian cell nucleus. Nature 404: 604–609. 2000. Dynamic associations of heterochromatin protein 1 Ragoczy, T., Telling, A., Sawado, T., Groudine, M., and Kosak, with the nuclear envelope. EMBO J. 19: 6558–6568. S.T. 2003. A genetic analysis of chromosome territory loop- Kurz, A., Lampel, S., Nickolenko, J.E., Bradl, J., Benner, A., Zir- ing: Diverse roles for distal regulatory elements. Chromo- bel, R.M., Cremer, T., and Lichter, P. 1996. Active and in- some Res. 6: 513–525. active genes localize preferentially in the periphery of chro- Ramalho-Santos, M., Yoon, S., Matsuzaki, Y., Mulligan, R.C., mosome territories. J. Cell. Biol. 135: 1195–1205. and Melton, D.A. 2002. “Stemness”: Transcriptional profil- Lachner, M., O’Carroll, D., Rea, S., Mechtler, K., and Jenuwein, ing of embryonic and adult stem cells. Science 298: 597–600. T. 2001. Methylation of histone H3 lysine 9 creates a binding Rea, S., Eisenhaber, F., O’Carroll, D., Strahl, B.D., Sun, Z.W., site for HP1 proteins. Nature 410: 116–120. Schmid, M., Opravil, S., Mechtler, K., Ponting, C.P., Allis, Lamond, A.I. and Spector, D.L. 2003. Nuclear speckles: A model C.D., et al. 2000. Regulation of chromatin structure by site- for nuclear organelles. Nat. Rev. Mol. Cell. Biol. 4: 605–612. specific histone H3 methyltransferases. Nature 406: 593– Leitch, A.R. 2000. Higher levels of organization in the inter- 599. phase nucleus of cycling and differentiated cells. Microbiol. Roix, J.J., McQueen, P.G., Munson, P.J., Parada, L.A., and Mis- Mol. Biol. Rev. 64: 138–152. teli, T. 2003. Spatial proximity of translocation-prone gene Lercher, M.J., Urrutia, A.O., and Hurst, L.D. 2002. Clustering of loci in human lymphomas. Nat. Genet. 34: 287–291. housekeeping genes provides a unified model of gene order in Rowley, J.D. 1998. The critical role of chromosome transloca- the human genome. Nat. Genet. 31: 180–183. tions in human leukemias. Annu. Rev. Genet. 32: 495–519. Leung, A.K. and Lamond, A.I. 2003. The dynamics of the Roy, P.J., Stuart, J.M., Lund, J., and Kim, S.K. 2002. Chromo- nucleolus. Crit. Rev. Eukaryot. Gene Expr. 13: 39–54. somal clustering of muscle-expressed genes in Caenorhab- Li, G., Sudlow, G., and Belmont, A.S. 1998. Interphase cell cycle ditis elegans. Nature 418: 975–979. dynamics of a late-replicating, heterochromatic homoge- Ruggero, D., Wang, Z.G., and Pandolfi, P.P. 2000. The puzzling neously staining region: Precise choreography of condensa- multiple lives of PML and its role in the genesis of cancer. tion/decondensation and nuclear positioning. J. Cell. Biol. Bioessays 22: 827–835. 140: 975–989. Sacco-Bubulya, P. and Spector, D.L. 2002. Disassembly of inter- Lloyd, D.J., Trembath, R.C., and Shackleton, S. 2002. A novel chromatin granule clusters alters the coordination of tran- interaction between lamin A and SREBP1: Implications for scription and pre-mRNA splicing. J. Cell. Biol. 156: 425–436. partial lipodystrophy and other laminopathies. Hum. Mol. Schubeler, D., Francastel, C., Cimbora, D.M., Reik, A., Martin, Genet. 11: 769–777. D.I., and Groudine, M. 2000. Nuclear localization and his- Lundgren, M., Chow, C.M., Sabbattini, P., Georgiou, A., tone acetylation: A pathway for chromatin opening and tran- Minaee, S., and Dillon, N. 2000. Transcription factor dosage scriptional activation of the human ␤-globin locus. Genes &

GENES & DEVELOPMENT 1383 Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Kosak and Groudine

Dev. 14: 940–950. chromosomal positioning of the epidermal differentiation Skok, J.A., Brown, K.E., Azuara, V., Caparros, M.L., Baxter, J., complex (EDC) in keratinocyte and lymphoblast interphase Takacs, K., Dillon, N., Gray, D., Perry, R.P., Merken- nuclei. Exp. Cell Res. 272: 163–175. schlager, M., et al. 2001. Nonequivalent nuclear location of Zhong, S., Salomoni, P., and Pandolfi, P.P. 2000. The transcrip- immunoglobulin alleles in B lymphocytes. Nat. Immunol. tional role of PML and the nuclear body. Nat. Cell. Biol. 2: 848–854. 2: E85–E90. Spellman, P.T. and Rubin, G.M. 2002. Evidence for large do- Zirbel, R.M., Mathieu, U.R., Kurz, A., Cremer, T., and Lichter, mains of similarly expressed genes in the Drosophila ge- P. 1993. Evidence for a nuclear compartment of transcription nome. J. Biol. 1: 5. and splicing located at chromosome domain boundaries. Swift, H. 1959. Studies on nuclear fine structure. Brookhaven Chromosome Res. 1: 93–106. Symp. Biol. 12: 134–152. Tanabe, H., Muller, S., Neusser, M., von Hase, J., Calcagno, E., Cremer, M., Solovei, I., Cremer, C., and Cremer, T. 2002. Evolutionary conservation of chromosome territory arrange- ments in cell nuclei from higher primates. Proc. Natl. Acad. Sci. 99: 4424–4429. Thummel, C.S. 2002. Ecdysone-regulated puff genes 2000. In- sect Biochem. Mol. Biol. 32: 113–120. Tollervey, D., Lehtonen, H., Carmo-Fonseca, M., and Hurt, E.C. 1991. The small nucleolar RNP protein NOP1 (fibrillarin) is required for pre-rRNA processing in yeast. EMBO J. 10: 573– 583. Trinh, L.A., Ferrini, R., Cobb, B.S., Weinmann, A.S., Hahm, K., Ernst, P., Garraway, I.P., Merkenschlager, M., and Smale, S.T. 2001. Down-regulation of TDT transcription in CD4+CD8+ thymocytes by Ikaros proteins in direct compe- tition with an Ets activator. Genes & Dev. 15: 1817–1832. Vermaak, D., Ahmad, K., and Henikoff, S. 2003. Maintenance of chromatin states: An open-and-shut case. Curr. Opin. Cell. Biol. 15: 266–274. Verschure, P.J., van Der Kraan, I., Manders, E.M., and van Driel, R. 1999. Spatial relationship between transcription sites and chromosome territories. J. Cell. Biol. 147: 13–24. Versteeg, R., van Schaik, B.D., van Batenburg, M.F., Roos, M., Monajemi, R., Caron, H., Bussemaker, H.J., and van Kam- pen, A.H. 2003. The human transcriptome map reveals ex- tremes in gene density, intron length, GC content, and re- peat pattern for domains of highly and weakly expressed genes. Genome Res. 13: 1998–2004. Visser, A.E., Jaunin, F., Fakan, S., and Aten, J.A. 2000. High resolution analysis of interphase chromosome domains. J. Cell. Sci. 113: 2585–2593. Volpi, E.V., Chevret, E., Jones, T., Vatcheva, R., Williamson, J., Beck, S., Campbell, R.D., Goldsworthy, M., Powis, S.H., Ra- goussis, J., et al. 2000. Large-scale chromatin organization of the major histocompatibility complex and other regions of human chromosome 6 and its response to interferon in in- terphase nuclei. J. Cell. Sci. 113: 1565–1576. Wallin, J.J., Gackstetter, E.R., and Koshland, M.E. 1998. Depen- dence of BSAP repressor and activator functions on BSAP concentration. Science 279: 1961–1964. Wallrath, L.L. 1998. Unfolding the mysteries of heterochroma- tin. Curr. Opin. Genet. Dev. 8 147–153. Walter, J., Schermelleh, L., Cremer, M., Tashiro, S., and Cremer, T. 2003. Chromosome order in HeLa cells changes during mitosis and early G1, but is stably maintained during sub- sequent interphase stages. J. Cell. Biol. 160: 685–697. Wang, H., Boisvert, D., Kim, K.K., Kim, R., and Kim, S.H. 2000. Crystal structure of a fibrillarin homologue from Metha- nococcus jannaschii, a hyperthermophile, at 1.6 A resolu- tion. EMBO J. 19: 317–323. Wang, J., Shiels, C., Sasieni, P., Wu, P.J., Islam, S.A., Freemont, P.S., and Sheer, D. 2004. Promyelocytic leukemia nuclear bodies associate with transcriptionally active genomic re- gions. J. Cell. Biol. 164: 515–526. Williams, R.R., Broad, S., Sheer, D., and Ragoussis, J. 2002. Sub-

1384 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on September 26, 2021 - Published by Cold Spring Harbor Laboratory Press

Form follows function: the genomic organization of cellular differentiation

Steven T. Kosak and Mark Groudine

Genes Dev. 2004, 18: Access the most recent version at doi:10.1101/gad.1209304

References This article cites 97 articles, 39 of which can be accessed free at: http://genesdev.cshlp.org/content/18/12/1371.full.html#ref-list-1

License

Email Alerting Receive free email alerts when new articles cite this article - sign up in the box at the top Service right corner of the article or click here.

Cold Spring Harbor Laboratory Press